fix(files): reuse a stored file by digest only for the same uploader #66

Merged

claude merged 1 commit from MEDECA-630-dedup-same-uploader into main

2026-10-03 22:31:08 +00:00

claude commented

2026-10-03 21:50:26 +00:00

Member

Summary

For MEDECA-630. MongoFileStore.store(name, Array[Byte], metadata) deduplicated by SHA-256 across all callers. It returned the existing file, which keeps its first uploader's uploadedBy. A Klient who uploads the same bytes as another user then gets a file not recorded as theirs, and MEDECA's "the caller uploaded this file" check refuses their own upload.

New rule: reuse a stored file only when its uploadedBy equals the uploader of this call; otherwise store a new file stamped with this call's uploader.

The uploader of the call is the uploadedBy after withUploader: the current user when there is one, else the value in metadata.
No uploader on both sides: the existing file is reused (earlier behaviour).
Stored file has an uploader, call has none (or the reverse): a new file is stored. This is the conservative choice; an anonymous call never gets a file recorded as someone else's.
Among several stored files with the same digest, the one with the matching uploader is reused.
The streaming and FileRepr store paths do not deduplicate and are unchanged. Existing deduplicated files keep their first uploader.

Same patch on release/0.1.x: the 0.1.x PR. MongoFileStore and its dependencies are identical on both lines.

0.1.25 is to carry this change together with the MEDECA-655 workflow change.

Tests

New MongoFileStoreIntegrationSpec (files/adapters/mongo/it) runs against a real MongoDB named by MONGO_URI (skipped when unset, as the existing spec in that module). It uses its own database per run and drops it at the end.

MONGO_URI=mongodb://localhost:<port> ./mill filesMongo.it.testOnly works.iterative.core.service.impl.MongoFileStoreIntegrationSpec against a disposable mongo:5 container: 7 passed, 0 failed. Before the fix: 3 passed, 4 failed (the 4 new-rule cases).
Pre-push ./mill __.test: passed. Note that __.test does not include filesMongo.it (its task is testForked), so CI does not run this spec.

Known pre-existing defect (not changed here): MongoJsonFileRepositoryIntegrationSpec in the same module connects to localhost:27017 even when MONGO_URI is set, because MongoClient.layer reads the config through the test environment, which is empty. It fails when nothing listens on 27017. The new spec reads the live MONGO_URI and avoids this.

🤖 Generated with Claude Code

## Summary For **MEDECA-630**. `MongoFileStore.store(name, Array[Byte], metadata)` deduplicated by SHA-256 across all callers. It returned the existing file, which keeps its first uploader's `uploadedBy`. A Klient who uploads the same bytes as another user then gets a file not recorded as theirs, and MEDECA's "the caller uploaded this file" check refuses their own upload. New rule: reuse a stored file only when its `uploadedBy` equals the uploader of this call; otherwise store a new file stamped with this call's uploader. - The uploader of the call is the `uploadedBy` after `withUploader`: the current user when there is one, else the value in `metadata`. - No uploader on **both** sides: the existing file is reused (earlier behaviour). - Stored file has an uploader, call has none (or the reverse): a new file is stored. This is the conservative choice; an anonymous call never gets a file recorded as someone else's. - Among several stored files with the same digest, the one with the matching uploader is reused. - The streaming and `FileRepr` store paths do not deduplicate and are unchanged. Existing deduplicated files keep their first uploader. Same patch on `release/0.1.x`: the 0.1.x PR. `MongoFileStore` and its dependencies are identical on both lines. **0.1.25 is to carry this change together with the MEDECA-655 workflow change.** ## Tests New `MongoFileStoreIntegrationSpec` (`files/adapters/mongo/it`) runs against a real MongoDB named by `MONGO_URI` (skipped when unset, as the existing spec in that module). It uses its own database per run and drops it at the end. - `MONGO_URI=mongodb://localhost:<port> ./mill filesMongo.it.testOnly works.iterative.core.service.impl.MongoFileStoreIntegrationSpec` against a disposable `mongo:5` container: 7 passed, 0 failed. Before the fix: 3 passed, 4 failed (the 4 new-rule cases). - Pre-push `./mill __.test`: passed. Note that `__.test` does not include `filesMongo.it` (its task is `testForked`), so CI does not run this spec. **Known pre-existing defect (not changed here):** `MongoJsonFileRepositoryIntegrationSpec` in the same module connects to `localhost:27017` even when `MONGO_URI` is set, because `MongoClient.layer` reads the config through the test environment, which is empty. It fails when nothing listens on 27017. The new spec reads the live `MONGO_URI` and avoids this. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

claude added 1 commit

2026-10-03 21:50:26 +00:00

fix(files): reuse a stored file by digest only for the same uploader

CI / Check Formatting (pull_request) Successful in 1m13s

Details

CI / Compile (pull_request) Successful in 3m10s

Details

CI / Check Linting (pull_request) Successful in 3m12s

Details

CI / Run Tests (pull_request) Successful in 3m29s

Details

1c4a46ed81

MongoFileStore.store(name, Array[Byte], metadata) returned any stored
file with the same SHA-256 digest. That file kept the metadata of its
first uploader, so a second caller who uploaded the same bytes got a
file that was not recorded as theirs. A check of "did this caller
upload this file" then refused the caller's own upload.

The store now reuses a file only when its stored `uploadedBy` equals
the uploader of this call (the current user, else `uploadedBy` in the
given metadata). No uploader on both sides keeps the earlier behaviour.
Otherwise it stores a new file stamped with this call's uploader.

The integration spec runs against a real MongoDB named by MONGO_URI.

For MEDECA-630. To be released in 0.1.25 with the MEDECA-655 change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude referenced this pull request

2026-10-03 22:29:58 +00:00

CI: a GitHub outage fails every job at the JDK download #68

claude merged commit 15ff2e8a8b into main

2026-10-03 22:31:08 +00:00

claude referenced this pull request from a commit

2026-10-03 22:31:10 +00:00

Merge pull request 'fix(files): reuse a stored file by digest only for the same uploader' (#66) from MEDECA-630-dedup-same-uploader into main